AI-based approach for detecting FDI attacks in load frequency control for centralized multi-area power systems

The Improvements in centralized power systems have transformed the power grid into the most complex smart grid, which is a prime example of cyber physical systems (CPS), vulnerable to various kinds of cyber threats. A major part of the power system, the load frequency control (LFC) which is implemented to regulate the power in tie-lines and synchronize the frequency to its nominal value, in particular, is vulnerable to multiple type of false data injection (FDI) attacks. This poses significant threats to system reliability, continuity and stability. In response, this paper proposes an AI/ML-based approach to detect FDI attacks in a multi-area networked renewable energy resources controlled and managed by a supervisory control and data acquisition (SCADA) mechanism. Primarily, an AI model for multi-area networks is implemented, training a Levenberg–Marquardt fast neural network (LMFNN) model using collected dataset on tieline power deviations, frequency aberrations, and active power load deviations in both areas. Afterward, two methods are simulated to identify FDI attacks in the LFC system. In the first technique, the output control signal of the LMFNN is compared with the actual plant output to detect residuals indicative of FDI attacks. The second technique employs an AI/ML-based classification technique with two labels: system under attack or no attack and successfully achieved 0.99 score for overall regression coefficient R. The precision of the proposed approach is demonstrated through simulation models conducted on two interconnected power systems, considering centralized power generation among solar power plant and thermal power plant.